HomeTech NewsArtificial IntelligenceA new (and impressive) AI image generator is available: StableDiffusion

A new (and impressive) AI image generator is available: StableDiffusion

Published on

- Advertisement -

We are witnessing live the evolution of AI-powered imagers. Not only are there more and more proposals of this type, but many of them are crossing the barrier of research, becoming available to all users. A little over a month ago, the open betas for Midjourney and DALL-E 2 were announced. Now it’s StableDiffusion’s turn.

You may not have heard of StableDiffusion, but it is a diffusion model capable of generating photorealistic images from any text developed by a startup called Stability AI together with researchers from the University of Heidelberg (Germany). The images generated by this alternative have an impressive level of detail, and are more similar to DALL-E 2 than to other proposals such as Midjourney, whose essence is more artistic and less realistic.

- Advertisement -

StableDiffusion, available to everyone

Like other proposals, StableDiffusion was trained on data from the Internet. In this case, LAION-Aesthetics was used, a set with millions of images filtered and classified by AI with which they taught the model to learn the associations between written concepts and images. From the company they affirm that although this technique is very effective, it is exposed to “social biases and unsafe content available on the network”, for which they ask that it be used responsibly.

After initially being available to project collaborators and select researchers, StableDiffusion is now accessible to everyone. The stable version is available through DreamStudio, a front-end and payment API. The good news is that registration is free, and when you log in you receive 200 credits for generating images. Although, yes, 1 credit does not always equal an image. Let’s see.

- Advertisement -

Upon entering DreamStudio you will find a simple and friendly interface. To generate an image you simply have to enter the desired text (in English) in the box labeled “I want to dram of…” and adjust the image width, height, and other rendering parameters. As you move the controls, the amount of credits you will have to pay will increase or decrease.

dream studio 2

- Advertisement -

In our test, for example, we asked for “a futuristic metropolis, with flying cars and glass skyscrapers”, with the settings you can see on the screenshot. DreamStudio has “listed” his work in 11 credits. It seemed good to us, we clicked on Dream and has thrown the top image. But this is not the only option available.

Stablediffusion 12

The company has also publicly released a demo that is much easier to use. In this case you just have to enter the text and click on Generate image. Here we have tried with “a sunny day with a house with a huge garden“, and the result has been quite realistic. As we can see, it is one more tool to unleash our creativity and evaluate the progress of these systems.

It should be noted that the creators of StableDiffusion say they will continue to work on improving the capabilities of the model, including the ability to remove unwanted results. Later they will release a version that can be run locally. Although yes, you will need at least graphics cards such as the Nvidia GeForce GTX 1660.

If it seems to you that there are many AI’s available and you don’t know where to start, we recommend you to see this interesting Twitter thread by the researcher Fabian Stelzer, who has bought the results of DALL-E 2, Midjourney and Stable Diffusion.

In Xataka | AIs have a problem: they are opaque and closed. BLOOM is the great open source project that wants to change everything

- Advertisement -

Latest articles

50 professions that did not exist 30 years ago

The Internet changed society a little less than 30 years ago, and that also...

EU Environment Commissioner urges more action to save biodiversity

It is an almost invisible crisis that threatens our food security, our health and...

More like this